01 - THE STANDARD

The mechanism stays private. The results do not.

A number without its configuration is a marketing claim, not evidence. Every result we publish carries its model, hardware, context length, and the seed it can be replayed from, measured against the same workload on a standard cache. Anything that has not cleared that bar is not on this site.

02 - HOW WE REPORT A NUMBER

Every number travels with its setup.

A footprint or a recall delta means nothing on its own. Read on the wrong hardware, at the wrong context length, against a baseline you never saw, it can say almost anything. So we never report a value alone.

Each result ships as a card: the metric, the configuration that produced it, the seed it replays from, and the baseline it is measured against. Change one field and it is a different result, reported separately.

FIG. 1 - RESULT CARD
SCHEMA
metric: <what was measured>
value: <result, with units>
model: <name, size, precision>
hardware: <accelerator, memory>
context.length: <tokens>
seed: <integer>
validity: <how the run was measured>
baseline: standard KV cache, same workload

Illustrative schema. The fields are fixed, the values come from the run.

03 - WHAT WE MEASURE

Four axes, reported the same way every time.

FOOTPRINT
The number you set, held.

The total field budget under a stated configuration, shown against a standard cache on the same workload, so the ceiling you chose is the ceiling you get.

RECALL
What survives at depth.

Attention to early-conversation content, measured across the run rather than asserted at the end of it. Reported against the same workload on a standard cache, at a stated context length, so the comparison is a measurement rather than a claim.

LATENCY
The cost of reading a field.

Read and write latency at each fidelity setting, on the hardware named in the result, so the trade is visible before you commit to it.

FAILURE MODES
Where it gives ground.

The conditions where SGF underperforms, published next to the wins, because a number you can trust is one that admits its own edges.

SCOPE: these are the axes. The published run above reports footprint and validity; recall and latency for your own workload come out of the pilot, each carrying the configuration that produced it.

04 - THE PUBLISHED RESULT

Run 149/006, every field filled.

The card on the right is a real result, not a format demonstration. One unbroken conversation of 1,900,000 tokens across 9,359 turns, on an unmodified off-the-shelf model, with the field footprint unchanged from 50,000 tokens onward. The run is measured live and retained in full; its seed fixes the input workload, so the same conversation can be measured against a standard cache. The scope note below states what the card does and does not claim.

FIG. 2 - PUBLISHED RESULT
RUN 149/006
metric
field footprint over conversation length
value
19.1 MB, unchanged from 50,000 to 1,900,000 tokens
model
Llama-3.2-1B-Instruct, unmodified
hardware
a single consumer-grade NVIDIA Blackwell GPU
context.length
1,900,000 tokens over 9,359 turns
seed
1493
validity
measured live, full run retained
baseline
standard KV cache, same workload

Measured values. Total deployed memory across this run was approximately 45 MB against approximately 62 GB for a standard KV cache at the same depth. The 19.1 MB above is the cache replacement and is flat across the run; total deployed memory stays bounded in the low tens of MB at any depth. The ratio grows with length and is stated at its depth. This run demonstrates topical coherence and continued attention to early-conversation content. It is not a claim of exact fact recall. The field runs on a single consumer-grade NVIDIA Blackwell GPU and Apple Silicon (Metal) — no datacenter accelerator required.

05 - WHERE SGF IS THE WRONG TOOL

The edges we publish next to the wins.

A bounded field is not a universal upgrade. In some workloads it earns you nothing — and in those it costs you nothing either, so leaving it on does no harm. There is also one thing this run does not demonstrate. We'd rather you know the edges before a pilot than discover them after.

Short contexts

When the whole conversation fits comfortably inside the cache, there is nothing to bound — a standard cache carries it verbatim. SGF earns you nothing here, and costs you nothing: you can leave it on and lose nothing until the conversation outgrows the cache.

Single-turn prompts

One question, one answer, no history to hold. The field is built for memory that accumulates; a prompt that never accumulates has nothing for it to preserve. No benefit, and no penalty.

Exact long-range fact recall

The published run demonstrates topical coherence and continued attention to early-conversation content at depth. It does not demonstrate verbatim retrieval of a specific value from ~1.6 million tokens ago. If your workload turns on retrieving one exact string from deep history, that is not what this result shows. This run was not designed to test recall — exact retrieval is a separate capability that was not exercised here.

A deliberately weak substrate

The published run uses a 1B model, chosen to stress the field rather than to flatter the output. Answer quality on that substrate is not the product and should not be read as a ceiling. It is a floor.

06 - WHAT WE PUBLISH, WHAT WE DO NOT

Open with the results, closed on the method.

Everything a buyer needs to judge SGF is public: the footprint, the recall, the latency, and the conditions where it is the wrong tool. How centers are formed, scored, and merged is the invention, and it is held rather than published. The line is deliberate, and it does not move.

PUBLISHED

Footprint, recall, and latency, each with the model, hardware, context length, and lens configuration that produced it.

PUBLISHED

Where SGF underperforms: short contexts, single-turn prompts, and the edges this run does not demonstrate.

UNDER NDA

Parameter-level behavior, integration internals, and the pilot report measured on your own workload.

PATENT PENDING

Center formation, merge scoring, and the pin policy, the mechanism behind every number.

07 - PILOT PROGRAM

Run SGF against your own workload.

A scoped four-week pilot: you bring one workload and a memory budget, we return a measured comparison against your current cache: footprint, recall, and the configuration that produced both.

We'll reply to book your discovery call and share the scope sheet and qualifying questions, no deck.

© 2026 SEMGRAVTHE MECHANISM STAYS PRIVATE. THE RESULTS DO NOT.